New Features
- Bias detection & mitigation: Receive bias warnings in every part of the RapidMiner platform including Turbo Prep, Model Simulator and more. When Studio thinks you have a column that could lead to model bias, you’ll receive a warning along with an in-platform callout that explains what it was triggered by.
- Streaming & IIOT advancements: Mix and match RapidMiner with Python in low latency (50-100ms) use-cases, such as scoring large volumes of sensor data. Additionally, leverage a new function-fitting operator to fit data with custom functions when creating models for anomaly detection on devices, modeling physical behavior based on data, and more.
- Security enhancements: Support for Docker Rootless mode along with enhanced security in Kubernetes environments both raise our overall security standards. Security for containerized platforms is also improved through regular updates of Docker images with the newest secure components.
- Time series forecasting: Automate forecasting future values of univariate time series based on historical data in RapidMiner Go. Track advanced and seasonal trends when forecasting sales or staffing requirements and use intuitive visualizations to compare the results of competing models.
- NLP extension: Leverage a new RapidMiner extension for natural language processing to extract part-of-speech tags and recognize people, cities, organizations, and other entities within free text. This is typically used as a pre-processing method to determine the contents of documents, website text, etc.
Bugfixes
- The core Pivot operator now runs as expected inside a SparkRM operator.
- Updated heuristics for Hive table reads in Radoop Spark jobs to prevent failing Spark jobs when hidden Hive staging directories are present.